Add usage logging

[amontanez24]

Problem Description

It would be helpful for debugging and overall library management to have a better understanding of SDV usage. We want to log more information about how synthesizers are used.

Expected behavior

For each of the following events, create one INFO level log statement with what is required

Synthesizer Initialization

• Timestamp
• Synthesizer class name (GaussianCopula, CTGAN, HSA, etc.)
• Synthesizer id

Fit

• Timestamp
• Synthesizer class name (GaussianCopula, CTGAN, HSA, etc.)
• Statistics of the fit data:
  • Total # of tables
  • Total # of rows
  • Total # of columns
• Synthesizer Id

Sample

• Timestamp
• Synthesizer class name (GaussianCopula, CTGAN, HSA, etc.)
• Statistics of the sample size:
  • Total # of tables
  • Total # of rows
  • Total # of columns
• Synthesizer Id

Synthesizer save and load

• Timestamp
• Synthesizer class name (GaussianCopula, CTGAN, HSA, etc.)
• Synthesizer id

Metadata save

• Timestamp
• Metadata type (single or multi table)
• Statistics about the metadata
  • Total # of tables
  • Total # of columns
  • Total # of relationships

Additional context

We should use a different logger than the loggers we currently use to record this information.

Proposal

• Create a yaml file that defines the configuration for the logger that should be used for usage. Whenever we need the logger, we can load that yaml file and configure the logger from a dictionary.
• Store this yaml in sdv/logging
• Add a variable to the logger for how the logs should be stored (valid values will be None or 'local' for now). If set to None, we should not log this info.
  Example yaml

log_style: 'local'
version: 1
handlers:
  file:
    class: logging.FileHandler
    filename: sdv_logs.log
loggers:
  BasSynthesizer:
    level: INFO
    handlers: [file]
    propagate: no